基于FRS与GA-ELM的煤与瓦斯突出预测研究

摘要

In view of the complexity of the inner mechanism of the coal and gas outburst occurred,sudden factors and fuzziness between prominent events lead to the question of the prediction accuracy is not high,puts forward the fuzzy rough set theory (FRS) combined with improved extreme learning mechanism of the original data attribute di⁃mension,extract the cause of the important factors,as the ELM network input neurons,extreme learning machine us⁃ing genetic algorithm (GA) to optimize the hidden layer of network weights and threshold of GA-ELM prediction model is established,the model output for coal and gas outburst intensity forecast results. After experimental verifi⁃cation,the model generalization ability is well and high prediction accuracy.%针对煤与瓦斯突出发生内在机理复杂性、致突因素与突出事件之间模糊性导致预测精度不高这一问题，提出将模糊粗糙集理论（FRS）结合改进的极端学习机（ELM）进行煤与瓦斯突出预测。通过FRS信息约简理论降低致突因素原始数据属性维度，提取出致突辅助因素，与主要因素共同作为ELM网络神经元输入，利用遗传算法（GA）对极端学习机网络输入权值、隐含层阈值进行优化，建立GA-ELM预测模型，模型输出为煤与瓦斯突出强度预测结果。经过模型训练和试验验证，该模型泛化能力强、预测精度高、收敛速度明显加快。